Disciplines/fields: Cognitive Science, Computational Neuroscience

Duration: 4 sessions

Course Content

We will construct neural simulations that involve sensory processing, motor action, and cognition. Participants are encouraged to bring a laptop and follow along as we show how neurons can be organized to transform sensory stimuli into relevant motor actions. In particular, models involving classical conditioning and adaptive control will be presented.

In the first session, the Neural Engineering Framework will be introduced, which is a general method for implementing algorithms using biologically realistic neural models. We will also introduce Nengo, our open-source software toolkit for generating and interacting with these models. In the second session, we will explore a variety of models implemented with this system, focusing on simple perception, motor control, and memory tasks. The third session will examine cognitive control, showing how the basal ganglia and thalamus can be used to create flexible neural systems capable of responding appropriately in a variety of situations. Finally, the fourth session will investigate online learning, showing how neural connections can be modified based on experience to improve performance.

No prior experience with Nengo and the Neural Engineering Framework is assumed. For those that attended a similar course in IK 2015, this time there will be much more emphasis on learning and control.

Objectives

Conceptual: To understand the class of algorithms that neurons are well-suited to implement, and how these algorithms differ from standard algorithms

Methodological: To work with the software toolkit Nengo to quickly generate neural models that could be used to test theories of perception, cognition, and action.

Literature

Chris Eliasmith, Terrence C. Stewart, Xuan Choo, Trevor Bekolay, Travis DeWolf, Yichuan Tang, and Daniel Rasmussen. A large-scale model of the functioning brain. Science, 338:1202-1205, 2012. URL: http://nengo.ca/publications/spaunsciencepaper, doi:10.1126/science.1225266. 
T.C. Stewart and C. Eliasmith. Large-scale synthesis of functional spiking neural circuits. Proceedings of the IEEE, 102(5):881-898, May 2014. URL: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6779629, doi:10.1109/JPROC.2014.2306061

Vita

Terry Stewart is a Research Associate with the Centre for Theoretical Neuroscience at the University of Waterloo. Starting in engineering, his interest in cognitive function led him to a PhD in Cognitive Science at Carleton University, Canada, with a thesis on the philosophy of computational modelling. Since then, he has worked with Chris Eliasmith at the Centre for Theoretical Neuroscience building computational models that exhibit both cognitive behaviour and neurological accuracy. His core interest is in using these detailed computational models to evaluate high-level cognitive theories.

Link

http://terrystewart.ca